INDEX

Explanations

phrases prompting to check out something like articles or videos

oai_token-act-pair · gpt-3.5-turbo

phrases that encourage checking out or exploring additional content

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

enei

-0.73

 dictate

-0.71

 awake

-0.66

 fabrication

-0.65

ajor

-0.64

 rightful

-0.63

ylum

-0.63

 conviction

-0.63

 grasped

-0.63

Fourth

-0.62

POSITIVE LOGITS

 whats

0.81

nels

0.68

casts

0.67

how

0.65

icles

0.65

posts

0.63

www

0.62

flows

0.62

 Braun

0.61

fitted

0.61

Activations Density 0.024%

phrases prompting to check out something like articles or videos

phrases that encourage checking out or exploring additional content

No Comments

No Known Activations